CHANGES.txt

Changes
=======

Unreleased
----------

- ``flatten`` now only includes leaf nodes in the flattened dict.

- ``flatten`` does not include a path element for the name of the type node
  for sequences.

- ``unflatten`` is implemented.

- Added ``__setitem__`` to ``SchemaNode``, allowing replacement of nodes by
  name.

- Added ``get_value`` and ``set_value`` methods to ``Schema`` which allow
  access and mutation of appstructs using dotted name paths.

0.9.3 (2011-06-23)
------------------

- Add ``Time`` type.

- Add Dutch translation.

- Fix documentation: 0.9.2 requires ``deserialize`` of types to explicitly
  deal with the potential to receive ``colander.null``.

- Use ``default_tzinfo`` when deserializing naive datetimes.  See
  https://github.com/Pylons/colander/issues#issue/5.

- Allow ``default_tzinfo`` to be ``None`` when creating a
  ``colander.DateTime``.  See
  https://github.com/Pylons/colander/issues#issue/6.

- Add the ability to insert a ``colander.interfaces.Preparer`` between
  deserialization and validation. See the Preparing section in the
  documentation.

0.9.2 (2011-03-28)
------------------

- Added Polish translation, thanks to Jedrzej Nowak.

- Moved to Pylons Project GitHub (https://github.com/Pylons/colander).

- Add tox.ini for testing purposes.

- New API: ``colander.required``.  Used as the marker value when a
  ``missing`` argument is left unspecified.

- Bug fix: if a ``title`` argument which is the empty string or ``None`` is
  passed explicitly to a SchemaNode, it is no longer replaced by a title
  computed from the name.

- Add SchemaNode.__contains__ to support "name in schema".

- SchemaNode deserialization now unconditionally calls the schema type's
  ``deserialize`` method to obtain an appstruct before attempting to
  validate.  Third party schema types should now return ``colander.null`` if
  passed a ``colander.null`` value or another logically "empty" value as a
  cstruct during ``deserialize``.

0.9.1 (2010-12-02)
------------------

- When ``colander.null`` was unpickled, the reference created during
  unpickling was *not* a reference to the singleton but rather a new instance
  of the ``colander._null`` class.  This was unintentional, because lots of
  code checks for ``if x is colander.null``, which will fail across pickling
  and unpickling.  Now the reference created when ``colander.null`` is
  pickled is unpickled as the singleton itself.

0.9  (2010-11-28)
-----------------

- SchemaNode constructor now accepts arbitrary keyword arguments.  It
  sets any unknown values within the ``**kw`` sequence as attributes
  of the node object.

- Added Spanish locale:  thanks to Douglas Cerna for the translations!

- If you use a schema with deferred ``validator``, ``missing`` or
  ``default`` attributes, but you use it to perform serialization and
  deserialization without calling its ``bind`` method:

  - If ``validator`` is deferred, no validation will be performed.

  - If ``missing`` is deferred, the field will be considered *required*.

  - If ``default`` is deferred, the serialization default will be
    assumed to be ``colander.null``.

- Undocumented internal API for all type objects: ``flatten``.
  External type objects should now inherit from
  ``colander.SchemaType`` to get a default implementation.

0.8  (2010/09/08)
-----------------

- Docstring fixes to ``colander.SchemaNode`` (``missing`` is not the
  ``null`` value when required, it's a special marker value).

- The concept of "schema binding" was added, which allows for a more
  declarative-looking spelling of schemas and schema nodes which have
  dependencies on values available after the schema has already been
  fully constructed.  See the new narrative chapter in the
  documentation entitled "Schema Binding".

- The interface of ``colander.SchemaNode`` has grown a ``__delitem__``
  method.  The ``__iter__``, and ``__getitem__`` methods have now also
  been properly documented.

0.7.3 (2010/09/02)
------------------

- The title of a schema node now defaults to a titleization of the
  ``name``.  Underscores in the ``name`` are replaced with empty
  strings and the first letter of every resulting word is capitalized.
  Previously the ``name`` was not split on underscores, and the
  entirety of the ``name`` was capitalized.

- A method of the ``colander.Invalid`` exception named ``messages``
  was added.  It returns an iterable of error messages using the
  ``msg`` attribute of its related exception node.  If the ``msg``
  attribute is iterable, it is returned.  If it is not iterable, a
  single-element list containing the ``msg`` value is returned.

0.7.2 (2010/08/30)
------------------

- Add an ``colander.SchemaNode.__iter__`` method, which iterates over
  the children nodes of a schema node.

- The constructor of a ``colander.SchemaNode`` now accepts a
  ``widget`` keyword argument, for use by Deform (it is not used
  internally).

0.7.1 (2010/06/12)
------------------

- Make it possible to use ``colander.null`` as a ``missing`` argument
  to ``colander.SchemaNode`` for roundtripping purposes.

- Make it possible to pickle ``colander.null``.

0.7.0
-----

A release centered around normalizing the treatment of default and
missing values.

Bug Fixes
~~~~~~~~~

- Allow ``colander.Regex`` validator to accept a pattern object
  instead of just a string.

- Get rid of circular reference in Invalid exceptions: Invalid
  exceptions now no longer have a ``parent`` attribute.  Instead, they
  have a ``positional`` attribute, which signifies that the parent
  node type of the schema node to which they relate inherits from
  Positional.  This attribute isn't an API; it's used only internally
  for reporting.

- Raise a ``TypeError`` when bogus keyword arguments are passed to
  ``colander.SchemaNode``.

Backwards Incompatiblities / New Features
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

- ``missing`` constructor arg to SchemaNode: signifies
  *deserialization* default, disambiguated from ``default`` which acted
  as both serialization and deserialization default previously.

  Changes necessitated / made possible by SchemaNode ``missing``
  addition:

  - The ``allow_empty`` argument of the ``colander.String`` type was
    removed (use ``missing=''`` as a wrapper SchemaNode argument
    instead).

- New concept: ``colander.null`` input to serialization and
  deserialization.  Use of ``colander.null`` normalizes serialization
  and deserialization default handling.

  Changes necessitated / made possible by ``colander.null`` addition:

  - ``partial`` argument and attribute of colander.MappingSchema has
     been removed; all serializations are partial, and partial
     deserializations are not necessary.

  - ``colander.null`` values are added to the cstruct for partial
     serializations instead of omitting missing node values from
     the cstruct.

  - ``colander.null`` may now be present in serialized and
     deserialized data structures.

  - ``sdefault`` attribute of SchemaNode has been removed; we never need
    to serialize a default anymore.

  - The value ``colander.null`` will be passed as ``appstruct`` to
    each type's ``serialize`` method when a mapping appstruct doesn't
    have a corresponding key instead of ``None``, as was the practice
    previously.

  - The value ``colander.null`` will be passed as ``cstruct`` to
    each type's ``deserialize`` method when a mapping cstruct
    doesn't have a corresponding key instead of ``None``, as was the
    practice previously.

  - Types now must handle ``colander.null`` explicitly during
    serialization.

- Updated and expanded documentation, particularly with respect to new
  ``colander.null`` handling.

- The ``value`` argument`` to the ``serialize`` method of a SchemaNode
  is now named ``appstruct``.  It is no longer a required argument; it
  defaults to ``colander.null`` now.

  The ``value`` argument to the ``deserialize`` method of a SchemaNode
  is now named ``cstruct``.  It is no longer a required argument; it
  defaults to ``colander.null`` now.

- The ``value`` argument to the ``serialize`` method of each built-in
  type is now named ``appstruct``, and is now required: it is no
  longer a keyword argument that has a default.

  The ``value`` argument to the ``deserialize`` method of each
  built-in type is now named ``cstruct``, and is now required: it is
  no longer a keyword argument that has a default.

0.6.2 (2010-05-08)
------------------

- The default ``encoding`` parameter value to the ``colander.String``
  type is still ``None``, however its meaning has changed.  An
  encoding of ``None`` now means that no special encoding and decoding
  of Unicode values is done by the String type.  This differs from the
  previous behavior, where ``None`` implied that the encoding was
  ``utf-8``.  Pass the encoding as ``utf-8`` specifically to get the
  older behavior back.  This is in support of Deform.

- The default ``err_template`` value attached to the ``colander.Date``
  and ``colander.Datetime`` types was changed.  It is now simply
  ``Invalid date`` instead of ``_('${val} cannot be parsed as an
  iso8601 date: ${err}')``.  This is in support of Deform.

- Fix bug in ``colander.Boolean`` that attempted to call ``.lower`` on
  a bool value when a default value was found for the schema node.

0.6.1 (2010-05-04)
------------------

- Add a Decimal type (number type which uses ``decimal.Decimal`` as a
  deserialization target).

0.6.0 (2010-05-02)
------------------

- (Hopefully) fix intermittent datetime-granularity-related test
  failures.

- Internationalized error messages.  This required some changes to
  error message formatting, which may impact you if you were feeding
  colander an error message template.

- New project dependency: ``translationstring`` package for
  internationalization.

- New argument to ``colander.String`` constructor: ``allow_empty``.
  This is a boolean representing whether an empty string is a valid
  value during deserialization, defaulting to ``False``.

- Add minimal documentation about the composition of a
  colander.Invalid exception to the narrative docs.

- Add (existing, but previously non-API) colander.Invalid attributes
  to its interface within the API documentation.

0.5.2 (2010-04-09)
------------------

- Add Email and Regex validators (courtesy Steve Howe).

- Raise a ``colander.Invalid`` error if a ``colander.SequenceSchema``
  is created with more than one member.

- Add ``Function`` validator.

- Fix bug in serialization of non-Unicode values in the ``String`` class.

- Get rid of ``pserialize`` in favor of making ``serialize`` always
  partially serialize.

- Get rid of ``pdeserialize``: it existed only for symmetry.  We'll
  add something like it back later if we need it.

0.5.1 (2010-04-02)
------------------

- The constructor arguments to a the ``colander.Schema`` class are now
  sent to the constructed SchemaNode rather than to the type it represents.

- Allow ``colander.Date`` and ``colander.DateTime`` invalid error
  messages to be customized.

- Add a ``pos`` argument to the ``colander.Invalid.add`` method.

- Add a ``__setitem__`` method to the ``colander.Invalid`` class.

- The ``colander.Mapping`` constructor keyword argument
  ``unknown_keys`` has been renamed to ``unknown``.

- Allow ``colander.Mapping`` type to accept a new constructor
  argument: ``partial``.

- New interface methods required by types and schema nodes:
  ``pserialize`` and ``pdeserialize``.  These partially serialize or
  partially deserialize a value (the definition of "partial" is up to
  the type).

0.5 (2010-03-31)
----------------

- 0.4 was mispackaged (CHANGES.txt missing); no code changes from 0.4
  however.

0.4 (2010-03-30)
----------------

- Add ``colander.DateTime`` and ``colander.Date`` data types.

- Depend on the ``iso8601`` package for date support.

0.3 (2010-03-29)
----------------

- Subnodes of a schema node are now kept in the ``children`` attribute
  rather than the ``nodes`` attribute.

- Add an ``sdefault`` property to ``colander.SchemaNode`` objects.

- Add a ``clone`` method to ``colander.SchemaNode`` objects.

- Add a ``__str__`` method to the ``colander.Invalid`` exception that
  prints an error summary.

- Various error message improvements.

- Add ``colander.Length`` validator class.

0.2 (2010-03-23)
----------------

- Make nodetype overrideable.

- Add __getitem__ to SchemaNode.

- Fix OneOf message.

- Capitalize node titles.

- Deal with empty strings in String, Boolean, and Float types.

- Introduce description; make title the same as name.

- Remove copy method from schemanode.

- Allow schema nodes to have titles.

- The term "structure" is too overloaded to use for schema purposes:
  structure -> schema node.

- Make Sequence more like Tuple and Mapping (it uses a substructure
  rather than a structure parameter to denote its type).

- Add __repr__ and copy methods to structure.

- Add accept_scalar flag to Sequence.


0.1 (2010-03-14)
----------------

- Initial release.